Finseo is now a connector in Claude, and your AI visibility data comes with it
The Finseo Claude connector puts your AI search visibility data inside Claude: fifteen read only tools, connected with your Finseo login, no API key and no JSON config.
It is for marketing teams and agencies who already work in Claude and are tired of exporting a dashboard into a chat window to ask one question about it. This article covers what the connector does, what the directory label means, what it cannot do, and the data sitting behind it.
What went live
Finseo is listed in Anthropic's Connectors Directory as a Community connector, added in July 2026, under the categories Data, Sales and Marketing, and Communication. The endpoint is https://api.finseo.ai/v1/mcp, and it is the same endpoint that already served Finseo's MCP server to Cursor and Claude Code.
What changed is discovery and setup. Before, you added Finseo as a custom connector by pasting a URL. Now you find it in the directory inside Claude and connect it in one step. The connection runs over OAuth 2.1 with Dynamic Client Registration and PKCE, so Claude registers itself and you sign in with the Finseo account you already have.
The same catalog serves Claude.ai, Claude Desktop, Claude mobile, Claude Code and Cowork, which means one connection covers every surface your team uses.
The fifteen tools, and what each one answers
Every tool is read only. Claude can analyse your data and can never change it.
| Tool | The question it answers |
|---|---|
list_projects | Which projects and models am I tracking? |
get_visibility_metrics | What are my visibility, mention and citation rates, versus the period before? |
get_visibility_timeseries | How did those rates move, day by day? |
list_prompts | Which tracked prompts perform, and which do not? |
get_prompt_details | For one prompt, what happened per model and per answer? |
get_answer_content | What did the model literally write on this day, and what did it cite? |
get_competitor_ranking | Who is ahead of me, on which KPI? |
get_competitor_gap_analysis | Which answers name them and not me, and on which prompts? |
get_competitor_h2h | Who gets named first, and what do models claim about each of us? |
get_sentiment_overview | What is my Share of Voice, and what is said negatively about me? |
get_top_sources | Which sources do models cite in my category? |
get_tags | How is my prompt set segmented? |
get_query_fanouts | Which sub-queries do engines actually run? |
list_tasks | What is on my optimisation backlog, by priority? |
get_task_details | For the top task, what exactly do I do, and on what evidence? |
The four tools that matter most in practice are the drill-down chain, and it is the same one our competitor analysis in AI search runs on. get_competitor_ranking gives you a competitor id, get_competitor_gap_analysis takes that id and returns the prompts where only they appear, and get_answer_content returns the answer text that proves it. That sequence is three clicks in the dashboard and one sentence in Claude.
What "Community" on the listing means
Anthropic's directory carries two labels, and the distinction is worth stating plainly because it is ours to be honest about. Anthropic's own documentation says of the verified label: "Verification means Anthropic has reviewed the connector more closely than a Community connector, but it is not a security audit or a guarantee of how the connector will perform."
On community connectors it says: "Anthropic screens community connectors before listing, but has not reviewed this connector in depth."
Two consequences follow. First, the label is about review depth, not capability. Anthropic states that "once connected, a community connector has the same capabilities and access as any connector you grant". Second, the security question is one you should ask us, not Anthropic. Our answer: every tool is read only, tokens are scoped to your account, Bearer API keys are SHA-256 hashed before storage, and you can revoke the connection from Claude's settings at any time, which invalidates the issued tokens immediately.
One more mechanic matters commercially. Directory listings are eligible for Suggested Connectors, Anthropic's in chat recommendations, and Anthropic says directory ranking is "usage-based, similar to other app stores". Being in the directory is a distribution channel, not just a convenience.
Where this sits in a crowded directory
The directory is not small any more. It launched in July 2025, held more than 75 connectors in December 2025, more than 200 in April 2026, and 824 listings when we counted the live directory on 19 September 2026. Read those as directory supply, not as usage: Anthropic publishes no per connector install or invocation figure, and neither do we.
| Category | Example connector | What it brings into Claude |
|---|---|---|
| Classic SEO data | Semrush | 28.6B keywords, 808M domain profiles, 43T backlinks |
| SEO plus some AI answer data | Ahrefs | Rankings, backlinks, brand mentions in search and LLM answers |
| CRM and campaigns | HubSpot | Contacts, deals, campaigns, landing pages |
| Lifecycle automation | ActiveCampaign | 32 tools for lists, tags, email performance, automations |
| AI answer measurement | Finseo | Visibility, competitor gaps, cited sources, fan-outs, sentiment, tasks |
Semrush and Ahrefs are in the directory and both are good at what they measure. The distinction is the surface. They are built on the ten blue links and are adding AI answer data to that foundation. Our corpus is AI answers and nothing else, which is why the tool list reads the way it does: gap analysis per prompt, mention order, mention depth, the literal answer text and the sources an answer cites under it.
The data behind the tools
The connector is a thin layer. What it reads is the same corpus our AI visibility tracking runs on.
| What we hold | Rows, as of 1 September 2026 |
|---|---|
| Analysed AI answers | 8,938,509 |
| Cited sources | 8,846,011 |
| Source mentions | 69,967,000 |
| Competitor mentions | 64,294,891 |
| People Also Ask questions | 3,413,772 |
| Observed fan-out queries | 864,409 |
In the ninety days to 31 August 2026 that corpus took 4,201,041 new answers across eleven models. Your account sees only your own projects through the connector, never the aggregate, and never anybody else's data.
Two numbers from that corpus explain why a chat interface suits this data. Visibility rates differ enormously by model: in the fourteen days to 1 September 2026 the tracked brand was visible in 40.9% of Gemini answers and 21.8% of Claude answers, across 65,778 and 31,313 answers respectively. Visible here means named in the answer text or cited as a source, which is the same definition the get_visibility_metrics tool returns. A single headline number hides a spread that wide, and a conversation gets you to the split in one question. The second number is fan-out. We record 864,409 sub-queries that engines generated rather than users typing them, which is a data shape that only makes sense when you can ask follow up questions of it, and which we cover in detail in our work on query fan-out.
Method note
The corpus figures above are planner estimates of live row counts taken on 1 September 2026, not exact counts, because an exact count on a nine million row table costs minutes and the article rounds them anyway. The visibility rates come from answers in the fourteen days to that date, restricted to models with more than 500 answers in the window.
All of it is aggregate. No customer, project, private domain or account is identifiable in any figure here, and the connector itself never exposes aggregate data to an account.
The skew matters. Our corpus is made of prompts our customers chose to track, so it is large but not a random sample of AI search. Read the visibility rates as figures for brands of this kind, tracked on prompts of this kind, not as a measurement of AI search in general. The ChatGPT interface figure in particular rests on 2,503 answers, which is two orders of magnitude thinner than the AI Overviews figure next to it.
How to connect it
- In Claude, open Settings then Connectors.
- Search the directory for Finseo and click Connect.
- Sign in with your Finseo account and click Allow access.
- Ask a question.
On Team and Enterprise plans an organisation Owner enables the connector under Organization settings then Connectors first, after which members connect individually. Claude Code takes the same endpoint from the command line, with OAuth or with a Bearer key from your API settings.
Questions worth starting with:
- "How has my visibility changed in the last 30 days, and how much of it is being named versus being cited?"
- "Rank my competitors and run a gap analysis against the strongest one. Which prompts do they win?"
- "Show me what ChatGPT answered yesterday for my top prompt, with the sources it cited."
- "What are my open optimisation tasks, and what exactly should I do for the highest priority one?"
A Finseo account on a paid plan with at least one tracking project is the prerequisite. The connector reads what the platform already tracks, so it adds no data of its own.
What it does not do
It does not write. There is no tool that edits a project, adds a prompt or changes a setting, by design, because the failure mode of a write enabled analytics connector is an agent quietly corrupting the data you are trying to measure.
It does not replace the dashboard. Anything with a chart, a date picker or a side by side comparison is faster in the platform than in prose.
It does not remove the reason to check. Claude summarises what the tools return, and a summary of a number is still a summary. For anything going into a client report, pull the figure from the tool output rather than from the sentence around it.
FAQ
Is a Claude connector the same thing as MCP? Nearly. A connector is Anthropic's product name for an integration built on the Model Context Protocol, the open standard Anthropic published in November 2024 and donated to the Agentic AI Foundation in December 2025. Every connector in the directory is an MCP server. Not every MCP server is in the directory.
Can I use the Finseo Claude connector for free? The connector costs nothing. It reads a Finseo account, so it needs a paid Finseo plan with at least one tracking project behind it. On the Claude side, custom and directory connectors are available on Free, Pro, Max, Team and Enterprise plans.
Is it safe to connect? Every Finseo tool is read only, the connection is OAuth scoped to your account over HTTPS, and disconnecting in Claude invalidates the tokens immediately. The broader caution applies to any third party connector, including ours: a connector's developer controls which tools it exposes and can change them, so connect developers you trust and read the scopes on the consent screen.
Does Claude keep the data private? Anthropic states that raw connector content is not used for model improvement unless it is copied directly into the conversation. On our side the connector returns only the projects your account already has access to, and we log calls for reliability rather than content.
Does it work in Claude Code and Cowork? Yes. Anthropic's directory documentation says the same catalog serves Claude.ai, Cowork, Desktop, mobile and Claude Code, so one connection covers all of them. Claude Code can also add the endpoint directly with a Bearer key.
What if I already added Finseo as a custom connector? It keeps working. Anthropic notes that a connection made before a directory listing keeps using the endpoint it was installed with and shows under "Custom". To move onto the directory listing, remove it and add it again from the directory.
Which AI engines does the data cover?
The ones the platform tracks: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Copilot, Grok and more, filterable per tool with the same model filter as the dashboard. Claude visibility is tracked as an engine too, which is a pleasant symmetry: you can ask Claude how visible you are inside Claude.
Next step
If you run AI search for clients, the same endpoint works for every account you manage: see what agencies do with it.
Sources
Ours
- Finseo tracked corpus, live row counts on 1 September 2026: answers, sources, source mentions, competitor mentions, People Also Ask questions and fan-out queries
- Finseo corpus window, answers and distinct models in the ninety days to 31 August 2026
- Finseo visibility by model, answers in the fourteen days to 1 September 2026, models with more than 500 answers
External
- Anthropic, Connector verification, on the verified and community labels: https://claude.com/docs/connectors/verification
- Anthropic, Connectors directory, on Suggested Connectors, usage based ranking and directory coverage: https://claude.com/docs/connectors/directory
- Anthropic (25 November 2024), Introducing the Model Context Protocol: https://www.anthropic.com/news/model-context-protocol
- Anthropic (9 December 2025), Donating MCP to the Agentic AI Foundation, more than 10,000 active public MCP servers and 97M+ monthly SDK downloads: https://www.anthropic.com/news/donating-mcp
- Anthropic (23 April 2026), New connectors in Claude for everyday life, directory above 200 connectors: https://claude.com/blog/connectors-directory
- Anthropic Connectors Directory, count of live listings on 19 September 2026: 824. Earlier counts published by third parties for mid 2026 are lower, so read this as a directory count on one date rather than a growth curve: https://claude.com/connectors
- Semrush, HubSpot and ActiveCampaign listings in the Anthropic Connectors Directory, 2026, for the comparison table: https://claude.com/connectors
Google Search
Save Finseo as a preferred source to see our articles more often in Top Stories, AI Overviews and AI Mode.
